Tencent has open-sourced its latest translation model, Hunyuan-MT-7B, after a standout performance at the WMT2025 machine translation competition, where it won 30 out of 31 language categories.
We're excited to announce the the open-source release of Hunyuan-MT-7B, our latest translation model that just won big at WMT2025! 🚀🏆
— Hunyuan (@TencentHunyuan) September 1, 2025
Hunyuan-MT-7B is a lightweight 7B model that's a true powerhouse. It dominated the competition by winning 30 out of 31 language categories,… pic.twitter.com/eKUTlGXcyW
Hunyuan-MT-7B is a 7-billion-parameter model that punches far above its weight. Despite its smaller size, it outperformed larger models under strict open-source and public-data constraints. On the Flores200 benchmark, it delivered results on par with closed-source systems like GPT-4.1, making it one of the strongest open translation models available today.
Along with the Hunyuan-MT-7B, Tencent also released the Hunyuan-MT-Chimera-7B. It is said to be the first open-source integrated translation model. It takes outputs from multiple models and refines them for more accurate and professional translations, making it especially useful for specialised domains.
The main highlight of the Hunyuan-MT-7B is its efficiency, offering lightning-fast inference that allows users to process more translation requests on the same hardware. Its cost-effective design also gives it deployment flexibility, making it suitable for both powerful servers and smaller edge devices. On top of that, the model covers a wide linguistic range, supporting 33 languages along with five minority languages.
While there are lots of Chinese AI models that are open-sourced, for example, we all know about the DeepSeek, or the recently launched 20B Qwen-Image model from Qwen. However, Elon Musk-led xAI also open-sourced its Grok 2.5 model and promised to do the same with Grok 3 within six months.
That’s all for now. But if you want to get the next AI update delivered straight to your phone, then join us on WhatsApp.
